|Publication number||US7062444 B2|
|Application number||US 10/057,705|
|Publication date||Jun 13, 2006|
|Filing date||Jan 24, 2002|
|Priority date||Jan 24, 2002|
|Also published as||US20030139930|
|Publication number||057705, 10057705, US 7062444 B2, US 7062444B2, US-B2-7062444, US7062444 B2, US7062444B2|
|Inventors||Liang He, ChuanQuan Xie, Song Cui|
|Original Assignee||Intel Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (31), Non-Patent Citations (17), Referenced by (34), Classifications (8), Legal Events (3)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This application is related to co-pending patent application Ser. No. 10/057,161 entitled, “A Data Transmission System and Method for DSR Application over GPRS”, filed Jan. 24, 2002, which application is assigned to the assignee of the present application.
The present invention relates to distributed speech recognition (DSR) technology. More specifically, the present invention relates to the architecture design, implementation, and optimization of a DSR system.
Currently, computer-supported communication via the Internet has, in a short time, become an important determinant of social life and a driving force of economic development in all developed and developing countries. In this field, global markets with huge potential for development have formed within a very short time out of a rapidly evolving process of innovation.
In order to meet the requirements of the explosive growth of Internet technology, speech researchers have been putting a great deal of effort into integrating speech functions into Internet applications. For simple applications, voice playback and voice recording functions may be sufficient. Although computationally intensive, speech recognition functions are needed for complex applications such as voice access to the World Wide Web on a personal digital assistant or wireless phone. DSR can effectively balance computational loads and utilize the bandwidth of heterogeneous networks for voice recognition applications.
Generally, there are three alternative strategies in the design of DSR architectures. The first strategy is server-only processing, in which all processing is done at the server side and the speech signal is transmitted to the server either through the Internet by using speech coding or via a second channel like telephone. Because all the recognition processing takes place at the server side, this approach has the smallest computational and memory requirements on the client, thereby allowing a wide range of client machines to access the speech-enabled application. The disadvantage of this approach is that users cannot access applications through a low-bandwidth connection.
The second conventional strategy is client-only processing, in which most of the speech processing is done at the client side and the results are transmitted to the server. The typical advantages are that a high-bandwidth connection is not required and recognition can be based on high-quality speech, because the sampling and feature extraction takes place at the client side. This approach is also less dependent on the transmission channel and is therefore more reliable. However, the type of clients that the speech-enabled applications can support is significantly limited with this approach because the client must be powerful enough to perform the heavy computation that takes place in the speech recognition process.
The third strategy is a client-server approach. In the client-server DSR processing model, front-end processing is done at the client side. The speech features are then transmitted to the server and finally the processing of the speech decoding and language understanding is performed at the server side.
At present, the client-server based DSR approach has not been exploited. This approach requires a great deal of effort to integrate a client-server based DSR system into Internet and wireless communication applications.
The difficulty of implementing an efficient client-server DSR architecture relates to diverse handheld device designs, network infrastructure complexity, network gateway or server diversity, and unstructured information content. To make a client-server based DSR system happen in industry and the commercial market, the development efforts of many diverse areas are needed. As such, a comprehensive, practical, and efficient architecture for DSR in a client/server design has not been achieved in the conventional art.
The features of the invention will be more fully understood by reference to the accompanying drawings, in which:
In the following detailed description, numerous specific details are set forth in order to provide a thorough understanding of the invention. However, it will be appreciated by one of ordinary skill in the art that the present invention shall not be limited to these specific details.
A DSR system based on a client-server model is presented in
The client-server approach has many benefits as follows:
Based on such architecture of one embodiment of the present invention, the working scenario for accessing documents by speech through the Internet is provided as
As shown in
As shown in
The audio capturer 101 captures speech and can fulfill some robust functionality for a demanding environment. The audio capturer's 101 major functions include recording, sampling a speech signal, and performing voice active detection or end-point detection.
The feature extractor 102 processes the speech data for speech features. Fundamentally, the feature extractor 102 fulfills the functionalities of MFCC and vector quantization. The feature extractor 102 typically performs a short-time Fourier analysis and extracts a sequence of observation vectors (or acoustic vectors) X=[x1, x2 . . . xt]. Many choices exist for the acoustic vectors, but the cepstral coefficients have exhibited the best performance to date. The sequence of acoustic vectors can either be modeled directly or vector-quantized first and then modeled.
The data wrapper 103: As shown in
The interpreter 104 allocates tasks to the audio capturer 101, speech feature extractor 102, and data wrapper 103, which capture, process speech input data for features, and communicate with the server browser 211 of the DSR gateway 200, respectively. The interpreter 104 also receives display content which is organized in a document-card style. Each document contains several information cards. A user can navigate through the cards, display the card contents, and advance from one card to the next. Card navigation is controlled by the speech recognition results generated in the DSR gateway module 200 through events.
The respective components of the DSR client module 100 of one embodiment of the invention were described above. The following section provides a detailed description of the DSR gateway module 200.
As shown in
The server browser 210 interacts with the DSR client module 100 and the utility platform 230. The server browser 210 can be separated into three key components: the data wrapper 211, the Hypertext Transfer Protocol (HTTP) wrapper 212, and the interpreter 213.
When the server browser 210 receives the document content from the DSR document server 300 through the HTTP wrapper 212, the server browser 210 calls the interpreter 213 to parse the content and to allocate various tasks to the utility platform 230. These various tasks include generating display content for the client and grammar generation.
When display content has been generated in a document-card style as described above and the content has been sent to the server browser 210, the server browser 210 sends the content to the DSR client module 100 through the data wrapper 211 and 103. When the server browser 210 receives the speech features from the DSR client module 100, the server browser 210 forwards the speech features and allocates one more task to the utility platform 230 which performs speech recognition based on speech features and builds related grammar.
After speech recognition, the utility platform 230 sends the speech recognition results to the server browser 210. Finally, according to the speech recognition results, the server browser 210 sends a new display card identifier or a new document to the DSR client module 100.
The respective components of the server browser 210 are described in more detail below. The server browser 210 includes the data wrapper 211, the HTTP wrapper 212, and the interpreter 213.
Data wrapper 211 is similar to data wrapper 103 in the DSR client 100. Data wrapper 211 fulfills the jobs of receiving speech features from and sending display contents and card identifiers to the DSR client 100. Specifically, data wrapper 211 provides among the following functions: adjusting transmission control conditions according to transmission parameters; implementing the function of the TCP server; sending DSRML data and receiving an initial connection request; controlling whether or not to decompose a QoS header of received data packets according to a QoS control parameter; performing multi-frame synchronization; performing error detection and mitigation; performing feature decompression; receiving UDP packets and unwrapping them into proper formats; and receiving TCP packets and unwrapping them into proper formats.
The HTTP wrapper 212 is responsible for DSRML document requests and document transmission between the DSR document server 300 and the DSR server browser 210. This transmission is implemented using a standard HTTP conversation in one embodiment.
The interpreter 213 is the parser of DSRML. After receiving a DSRML document from the DSR document server 300, the interpreter 213 will parse the elements of a DSRML document, determine the syntax of the DSRML document, consolidate the information for the server browser 210, and assign various tasks to the utility platform 230.
The server browser 210, according to one embodiment of the invention was described above. The other two parts of the DSR gateway module 200 are described below. These other two parts include the DNS server 220 and the utility platform 230.
The DNS server 220 is implemented within the DSR gateway module 200. Using the DNS Server 220, the HTTP wrapper 212 of the DSR server browser 210 can easily and rapidly get a domain name interpretation service. The DNS server 220 caches the IP addresses of DSRML document servers while getting a domain name interpretation service from the external DNS servers.
The utility platform 230 is controlled by the server browser 210, and fulfills most of the low-level system jobs. The utility platform 230 can be separated into seven parts: the utility platform shell 231, the dynamic grammar builder 232, the transmission controller 233, the display content generator 234, the DSR speech recognition engine 235, the Text to Speech (TTS) engine 236, and the workload balance controller 237.
The utility platform shell 231 receives the various tasks from the server browser 210, allocates the tasks to specific components of the utility platform 230, and harmonizes the operation of all of the utility platform 230 components. Finally, the utility platform 231 communicates the results to the server browser 210.
The dynamic grammar builder 232 processes grammar from the speech input. In a DSR application each dialog has its own grammar, which specifies the allowable input word. Usually, the documents provide grammar as the text format of keywords without garbage from natural speech. Further, the speech recognition engine needs the compiled binary grammar to process the speech recognition. In particular, the dynamic grammar builder 232 performs among the following tasks:
extracting grammar scripts and keywords from documents.
adding garbage words for natural speaking input.
compiling the modified grammar to binary format for DSR speech recognition.
putting the binary format grammar files in cache.
The grammar builder 232 could be implemented as a tool in the speech recognition engine 235. These two components should be supplied and upgraded together, but there are many benefits if these components are separated into two components as described herein. These benefits include:
The grammar can only be compiled once and saved in cache for multiple users' threads;
Besides compiling for binary, the grammar builder 232 needs to deal with DSRML, cache, etc. Separate components can facilitate architectural design, implementation, and system management.
The display content generator 234 consolidates the display contents in DSRML documents and organizes the content into a document-card model like wireless markup language (WML), which matches with the dialog style in a DSRML document. In this manner, each dialog just matches one card with a corresponding card identifier. DSRML documents enable the interaction between the client and server using a speech input approach. The DSRML document contains the grammar script for speech recognition and the corresponding display information.
The display contents include DSRML script and other external resources such as pictures and grammar. This content is packed together and sent to the client through the network.
The transmission controller 233 mainly defines the communication model between the DSR client module 100 and the server browser 210 of the DSR gateway module 200. This component enables the use of diverse networks and protocols. The main functions of this component are to detect client type, to balance network load, and to adjust the transmission performance. Some control parameters are sent to the client at the session start. These parameters only need to be sent once. The other parameters are sent during the session, when network status changes. The control parameters might be, for example:
CRC: whether the transmission needs CRC, and what kind of CRC to use
Bandwidth: what are the bandwidth requirements for the feature transmission
Frame number: according to a particular network status, the proper number of frames to form a multi-frame according to network status.
QoS related parameters: whether a QoS header should be wrapped. Specifying the kind of QoS parameters.
These control parameters can be sent to clients in a special defined markup language format by event. For instance,
The parameters sent at session start could include, for example:
The parameters sent during the session could include, for example:
< framnum> 16 </framnum>
The workload balance controller 237 balances processing and resource needs of the other components of the utility platform 230. The DSR gateway 200 platform consumes a large amount of power. If the capacity requirement for current users is high, it can be beneficial to balance the workload using clustering servers. In the workload balance controller 237 of one embodiment, five sub-controllers are defined,
The DSR speech recognition engine 235 does the work of speech recognition from speech feature extractions. As the DSR speech recognition engine 235 needs binary format grammar, the dynamic grammar builder 232 must be called first.
The TTS engine 236 expands the system to support text input. Sometimes, the display approach in audio format is efficient and user preferred. As such, TTS technology is supported in the present invention.
The components of one embodiment of a comprehensive, practical, and efficient architecture for a DSR client and server development platform were described above by reference to
In the DSR development platform of the present invention, events are used to control all the interactions between the DSR client module 100 and the DSR gateway module 200. Two fundamental types of events are defined for the synchronization between the DSR client module 100 and the DSR gateway module 200. The events include:
system synchronization events (for normal and abnormal cases).
content synchronization events (only for abnormal cases).
System Synchronization Events
System synchronization events are defined to inform the client of what has happened and of what to do next. The system synchronization events are sent from the DSR gateway 200 to the DSR client 100 in response to triggers or circumstances including the following:
client doesn't respond within the timeout interval.
client doesn't respond intelligibly.
client requests help.
inform the client to move on to next dialog.
error occurred (such as semantic error in the document).
inform the client to exit the interaction.
According to the above triggers or circumstances, the corresponding types of system synchronization events include the following:
>The system synchronization event types in normal cases:
Ready: The server is ready to receive input.
Forward: The message for the next display card ID.
>The system synchronization event types in abnormal cases:
Noinput: Default message.
Nomatch: Default message.
Help: Default message.
Error: Default error message. In this case, the DSR gateway 200 will exit.
Exit: Default message. In the case, the DSR gateway 200 will exit.
The system synchronization events are generated at the DSR gateway 200 and organized as markup language format. For example, when the DSR gateway 200 finishes the recognition of one input, the DSR gateway 200 will transit to a next dialog and inform the DSR client 100 to show the corresponding next card. Thus, the “forward” synchronization type of event will be sent to the DSR client 100, and the related display card ID will be sent through a <message>element. An example of this type of event is as follows:
<event type=“forward”> <message> #city </message> </event>
Content Synchronization Events
Content synchronization events are used to deal with the abnormal cases according to the procedure pre-defined in DSRML script to cope with the abnormality. These types of content synchronization events include:
Help: The messages for the help for which the user asked.
Noinput: The message for no response within the timeout interval.
Nomatch: The message for no correct recognition result.
Error: The message for some other error related to application execution.
The content synchronization events are interpreted by the gateway DSR 200. For instance, in the case of a wrong recognition result during the interaction, the DSR gateway 200 may transfer to another dialog and inform the client to show the related card. Thus, based on the content synchronization event, the system synchronization event “forward” will be generated and sent to the DSR client 100. Following are example scripts.
The script of the nomatch content synchronization event that will be interpreted by the DSR gateway 200 as follows:
The script of the nomatch system synchronization event generated by the DSR gateway 200 according to the script of a content synchronization event as follows:
As shown in
As shown in
In the DSR application system of the present invention, the interaction model between the client and gateway is voice and visual through heterogeneous networks. As such, the definition of the DSR markup language that provides voice-visual interaction between clients and the gateway is important. An embodiment of this DSRML definition of the present invention is provided below. In general, the DSRML should provide functionality to recognize spoken input, navigate among related document-cards, and provide speech synthesis for displaying content in audio format.
The DSRML language of one embodiment of the present invention provides a means for collecting spoken input, assigning the input to document-defined request variables, and making decisions as to how to execute the DSRML documents based on the user's input. A document may be linked to other documents through Universal Resource Identifiers (URIs).
The Consideration of Client Device Types
In one embodiment, DSRML is designed to meet the constraints of a wide range of small, narrowband client devices. These client devices are primarily characterized in some of the following ways:
Display size—small screen size and low resolution. A small mobile device such as a phone may only have a few lines of textual display, each line containing 8–12 characters.
Input devices—voice input plus other limited or special-purpose input methods. The phone typically has a numeric keypad and a few additional function-specific keys. The more sophisticated device may have software-programmable buttons, but may not have a mouse or other pointing device.
Computational resources—low power CPU and small memory size; often limited by power constraints.
Narrowband network connectivity—low bandwidth and high latency. Devices with 300 bps to 10 Kbps network connections and 5–10 second round-trip latency are not uncommon.
The Consideration of the Definition for DSR Markup Language
A DSRML document forms a conversational finite state machine. The user is always in one conversational state, or dialog, at a time. Each dialog determines the next dialog to which the user transitions.
In one embodiment of the present invention, there are two kinds of dialogs: link and submit. If the dialog's type attribute is “submit”, usually it has a <submit> element. The dialog is used to collect values for each of several <input> item variables.
The <dialog> may specify a grammar that defines the allowable word for the input. The user must provide values for <input> variables before proceeding to the <submit> element in the dialog.
Another dialog type is “link”. In this dialog, there is no specified <input> element; but there are links in the display contents. The dialog is used to transit to another dialog through the links. Each link may also specify a grammar.
An application is a set of documents sharing the same application root document. Whenever the user interacts with a document in an application, its application root document remains loaded while the user is navigating between other documents in the same application. The root document is unloaded when the user navigates to a document that is not in the application. While it is loaded, the application root document's grammars can be set to remain active during the application.
Each dialog has one or more speech grammars associated with it. In server browser directed applications, each dialog's grammars are active only when the user is in that dialog. Both the <input> element and the <link> element need at least one grammar.
In a typical DSRML application, the documents contain text and pictures as the display contents. The display content in each dialog is organized as a card showing in the mobile client. Each dialog corresponds to one card, and vice versa. There is a specific element named “display” containing all the visual contents.
In a typical DSRML application, the system may generate events, including nomatch, time expiration, and user requested help. By default, the system handles all these events, not needing to define them in a document. But, if we have defined them in a DSRML document as content synchronization events, the defined element, such as <event> and/or <help>, will be executed, for example, sending a warning message to the client.
The DSR platform of the present invention provides several advantages over the prior art. Specifically, software vendors can develop better components and tools for the platform using the architecture definition provided herein. Client manufacturers can use the components in the client platform to independently develop DSR handheld devices; Internet Service Providers (ISP) or telecom operators can use the definition and develop components for the server platform to integrate with the DSR gateway; and ICP (Internet content provider) can use the DSR-XML definition and edit tools in the platform to develop DSR service content.
In one aspect of the present invention, a DSR system development platform based on client-server framework, includes:
In another aspect of the invention, a method for accessing documents by speech in a DSR system based on the above described platform, includes the components:
Thus, an inventive DSR client and server development platform including a DSR client module, DSR gateway module, and DSR document server is disclosed. The scope of protection of the claims set forth below is not intended to be limited to the particulars described in connection with the detailed description of presently preferred embodiments. Accordingly, the present invention provides a comprehensive, practical, and efficient architecture for a DSR client and server development platform.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US6018710 *||Dec 13, 1996||Jan 25, 2000||Siemens Corporate Research, Inc.||Web-based interactive radio environment: WIRE|
|US6073100||Mar 31, 1997||Jun 6, 2000||Goodridge, Jr.; Alan G||Method and apparatus for synthesizing signals using transform-domain match-output extension|
|US6188985 *||Oct 3, 1997||Feb 13, 2001||Texas Instruments Incorporated||Wireless voice-activated device for control of a processor-based host system|
|US6195632||Nov 25, 1998||Feb 27, 2001||Matsushita Electric Industrial Co., Ltd.||Extracting formant-based source-filter data for coding and synthesis employing cost function and inverse filtering|
|US6226606||Nov 24, 1998||May 1, 2001||Microsoft Corporation||Method and apparatus for pitch tracking|
|US6269336 *||Oct 2, 1998||Jul 31, 2001||Motorola, Inc.||Voice browser for interactive services and methods thereof|
|US6404859||May 6, 1999||Jun 11, 2002||Lockheed Martin Corporation||Voice enabled system for remote access of information|
|US6594628 *||Apr 2, 1997||Jul 15, 2003||Qualcomm, Incorporated||Distributed voice recognition system|
|US6604075 *||Mar 14, 2000||Aug 5, 2003||Lucent Technologies Inc.||Web-based voice dialog interface|
|US6615171 *||Aug 13, 1999||Sep 2, 2003||International Business Machines Corporation||Portable acoustic interface for remote access to automatic speech/speaker recognition server|
|US6662163 *||Mar 30, 2000||Dec 9, 2003||Voxware, Inc.||System and method for programming portable devices from a remote computer system|
|US6738743||Mar 28, 2001||May 18, 2004||Intel Corporation||Unified client-server distributed architectures for spoken dialogue systems|
|US6748375||Sep 7, 2000||Jun 8, 2004||Microsoft Corporation||System and method for content retrieval|
|US6754200||Aug 25, 2000||Jun 22, 2004||Fujitsu Limited||Rate control system of TCP layer|
|US6801604||Jun 25, 2002||Oct 5, 2004||International Business Machines Corporation||Universal IP-based and scalable architectures across conversational applications using web services for speech and audio processing resources|
|US20020006124||Jan 5, 2001||Jan 17, 2002||Ray Jimenez||Methods and apparatus for an audio web retrieval telephone system|
|US20020022453||Mar 30, 2001||Feb 21, 2002||Horia Balog||Dynamic protocol selection and routing of content to mobile devices|
|US20020147579||Feb 2, 2001||Oct 10, 2002||Kushner William M.||Method and apparatus for speech reconstruction in a distributed speech recognition system|
|US20020184197||Aug 23, 2001||Dec 5, 2002||Intel Corporation||Information retrieval center|
|US20020184373||Mar 21, 2002||Dec 5, 2002||International Business Machines Corporation||Conversational networking via transport, coding and control conversational protocols|
|US20020194388||Dec 4, 2001||Dec 19, 2002||David Boloker||Systems and methods for implementing modular DOM (Document Object Model)-based multi-modal browsers|
|US20030093265||Nov 12, 2001||May 15, 2003||Bo Xu||Method and system of chinese speech pitch extraction|
|US20030139930||Jan 24, 2002||Jul 24, 2003||Liang He||Architecture for DSR client and server development platform|
|US20030161298||Feb 27, 2003||Aug 28, 2003||Janne Bergman||Multi-modal content and automatic speech recognition in wireless telecommunication systems|
|US20040056885||Sep 25, 2003||Mar 25, 2004||Fujitsu Limited||Multichannel information processing device|
|US20040057456||Sep 20, 2002||Mar 25, 2004||Liang He||Transmitting data over a general packet radio service wireless network|
|US20040199502||Apr 21, 2004||Oct 7, 2004||Microsoft Corporation||System and method for content retrieval|
|US20050059426||Aug 17, 2004||Mar 17, 2005||Ari Aarnio||Systems and methods for presenting and/or converting messages|
|WO2000058942A2 *||Mar 7, 2000||Oct 5, 2000||Koninkl Philips Electronics Nv||Client-server speech recognition|
|WO2001035389A1||Nov 10, 2000||May 17, 2001||Koninkl Philips Electronics Nv||Tone features for speech recognition|
|WO2001095312A1 *||May 2, 2001||Dec 13, 2001||Nokia Mobile Phones Ltd||Method and system for adaptive distributed speech recognition|
|1||Allman et al., "Increasing TCP's Initial Window", RFC 2414, Internet Engineering Task Force, Internet Draft, Sep. 1998.|
|2||Boersma, Paul; Accurate Short-Term Analysis Of The Fundamental Frequency And The Harmonics-To-Noise Ratio Of A Sampled Sound; Institute Of Phonetic Sciences, University of Amsterdam; Proceedings 17 (1993), pp. 97-110.|
|3||Cheng Zhang, et al., "The Study on DSR Transmissions Over GPRS", Intel China Ltd., 2002 IEEE, pp. II-2081 through II-2084, Intel Corporation, PRC.|
|4||Coyner et al., "Distributed speech recognition services (DSRS)", International Symposium on Multimedia Software Engineering, Proceedings, 11-13 Dec. 2000, pp. 59-66.|
|5||Distributed Speech Recognition-Aurora, Oct. 1, 2002, <http://www.etsi.org/technicalactiv/dsr/dsr.htm>, pp. 1-3.|
|6||Hermes, DIK J.; Measurement of pitch by subharmonic summation; J. Acoust. Soc. Am. 83 (1), Jan. 1988, (C) 1988 Acoustical Society of America, pp. 257-264.|
|7||Liu, Ph.D., Sharlene, et al.; The Effect of Fundamental Frequency on Mandarin Speech Recognition; 5<SUP>th </SUP>International Conference on Spoken Language Processing; 30<SUP>th </SUP>Nov.-4<SUP>th </SUP>Dec. 1998, Sydney, Australia, ICSLP '98 Proceedings, Th4R9, vol. 6, pp. 2647-2650.|
|8||Mathis et al., "TCP Selective Acknowledgement Options", RFC 2018, Internet Engineering Task Force, Internet Draft, Oct. 1996.|
|9||Pearce, David, Enabling New Speech Driven Services for Mobile Devices: An overview of the ETSI standards activities for Distributed Speech Recognition Front-ends, AVIOS 2000: The Speech Applications Conference, May 22-24, 2000, San Jose, CA, USA., <http://www.etsi.org/T<SUB>-</SUB>news/Documents/AVIOS DSR paper.pdf>, 12 pages.|
|10||Rabiner, Lawrence R., et al; A Comparative Performance Study of Several Pitch Detection Algorithms; IEEE Transactons On Acoustics, Speech, And Signal Processing, vol. ASSP-24, No. 5, Oct. 1976, pp. 399-418.|
|11||Search Report for PCT/US 02/35949, mailed Feb. 6, 2003, 2 pages.|
|12||Speech Processing, Transmission and Quality aspects (STQ); Distributed speech recognition; Front-end feature extraction algorithm; Compression algorithms, ETSI ES 201 108 V1.1.2 (Apr. 2000), ETSI Standard, (C) European Telecommunications Standards Institute 2000, F-06921 Sophia Antipolis Cedex-France, pp. 1-20.|
|13||WebSphere Transcoding Publisher, IBM(R) Products & Services> Software> Web Application Servers, http://www-3.ibm.com/software/webservers/transcoding/about.html, 4 pages.|
|14||Weiqi Zhang, et al., "The Study on Distributed Speech Recognition System", IEEE Procs, ICASSP, col. 3, Jun. 2000, Intel Architecture Development Lab, Lernout & Hauspie Asia Pacific, pp. 1431-1434.|
|15||Written Opinion for PCT/US 02/35949, mailed Oct. 23, 2003, 1 page.|
|16||Yeping Su, et al., "An Improved Cumulant-Based Blind Speech Separation Method", Intel Architecture Development Lab, 2000 IEEE, Intel Corporation, pp. 1867-1870.|
|17||Zhang et al., "The study on distributed speech recognition system", ICASSP '00, Proceedings, 5-9 Jun. 2000, vol. 3, pp. 1431-1434.|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7302390 *||Jan 8, 2003||Nov 27, 2007||Industrial Technology Research Institute||Configurable distributed speech recognition system|
|US7657424||Dec 3, 2004||Feb 2, 2010||Phoenix Solutions, Inc.||System and method for processing sentence based queries|
|US7672841||May 19, 2008||Mar 2, 2010||Phoenix Solutions, Inc.||Method for processing speech data for a distributed recognition system|
|US7702508||Dec 3, 2004||Apr 20, 2010||Phoenix Solutions, Inc.||System and method for natural language processing of query answers|
|US7725307||Aug 29, 2003||May 25, 2010||Phoenix Solutions, Inc.||Query engine for processing voice based queries including semantic decoding|
|US7725321||Jun 23, 2008||May 25, 2010||Phoenix Solutions, Inc.||Speech based query system using semantic decoding|
|US7873519||Oct 31, 2007||Jan 18, 2011||Phoenix Solutions, Inc.||Natural language speech lattice containing semantic variants|
|US7912702||Oct 31, 2007||Mar 22, 2011||Phoenix Solutions, Inc.||Statistical language model trained with semantic variants|
|US8024194 *||Dec 8, 2004||Sep 20, 2011||Nuance Communications, Inc.||Dynamic switching between local and remote speech rendering|
|US8099289 *||May 28, 2008||Jan 17, 2012||Sensory, Inc.||Voice interface and search for electronic devices including bluetooth headsets and remote systems|
|US8195467 *||Jul 10, 2008||Jun 5, 2012||Sensory, Incorporated||Voice interface and search for electronic devices including bluetooth headsets and remote systems|
|US8229734||Jun 23, 2008||Jul 24, 2012||Phoenix Solutions, Inc.||Semantic decoding of user queries|
|US8260619||Mar 30, 2009||Sep 4, 2012||Convergys Cmg Utah, Inc.||Method and system for creating natural language understanding grammars|
|US8275618 *||Oct 29, 2007||Sep 25, 2012||Nuance Communications, Inc.||Mobile dictation correction user interface|
|US8335690||Jan 17, 2012||Dec 18, 2012||Convergys Customer Management Delaware Llc||Method and system for creating natural language understanding grammars|
|US8412532 *||Nov 2, 2011||Apr 2, 2013||Google Inc.||Integration of embedded and network speech recognizers|
|US8635243||Aug 27, 2010||Jan 21, 2014||Research In Motion Limited||Sending a communications header with voice recording to send metadata for use in speech recognition, formatting, and search mobile search application|
|US8762152||Oct 1, 2007||Jun 24, 2014||Nuance Communications, Inc.||Speech recognition system interactive agent|
|US8838457||Aug 1, 2008||Sep 16, 2014||Vlingo Corporation||Using results of unstructured language model based speech recognition to control a system-level function of a mobile communications facility|
|US8868428 *||Aug 14, 2012||Oct 21, 2014||Google Inc.||Integration of embedded and network speech recognizers|
|US8880405||Oct 1, 2007||Nov 4, 2014||Vlingo Corporation||Application text entry in a mobile environment using a speech processing facility|
|US8886540||Aug 1, 2008||Nov 11, 2014||Vlingo Corporation||Using speech recognition results based on an unstructured language model in a mobile communication facility application|
|US8886545||Jan 21, 2010||Nov 11, 2014||Vlingo Corporation||Dealing with switch latency in speech recognition|
|US8949130||Oct 21, 2009||Feb 3, 2015||Vlingo Corporation||Internal and external speech recognition use with a mobile communication facility|
|US8949266||Aug 27, 2010||Feb 3, 2015||Vlingo Corporation||Multiple web-based content category searching in mobile search application|
|US8996379||Oct 1, 2007||Mar 31, 2015||Vlingo Corporation||Speech recognition text entry for software applications|
|US9076448 *||Oct 10, 2003||Jul 7, 2015||Nuance Communications, Inc.||Distributed real time speech recognition system|
|US20040249635 *||Mar 2, 2004||Dec 9, 2004||Bennett Ian M.||Method for processing speech signal features for streaming transport|
|US20050080625 *||Oct 10, 2003||Apr 14, 2005||Bennett Ian M.||Distributed real time speech recognition system|
|US20050086049 *||Dec 3, 2004||Apr 21, 2005||Bennett Ian M.||System & method for processing sentence based queries|
|US20110184740 *||Jul 28, 2011||Google Inc.||Integration of Embedded and Network Speech Recognizers|
|US20120084079 *||Apr 5, 2012||Google Inc.||Integration of Embedded and Network Speech Recognizers|
|US20120310645 *||Dec 6, 2012||Google Inc.||Integration of embedded and network speech recognizers|
|US20130041666 *||Aug 8, 2012||Feb 14, 2013||Samsung Electronics Co., Ltd.||Voice recognition apparatus, voice recognition server, voice recognition system and voice recognition method|
|U.S. Classification||704/275, 715/727, 704/E15.047, 704/E15.044|
|International Classification||G10L15/30, G10L15/26|
|Jan 24, 2002||AS||Assignment|
Owner name: INTEL CORPORATION, A CORPORATION OF DELAWARE, CALI
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:HE, LIANG;XIE, CHUANQUAN;CUI, SONG;REEL/FRAME:012543/0599;SIGNING DATES FROM 20020111 TO 20020115
|Dec 9, 2009||FPAY||Fee payment|
Year of fee payment: 4
|Nov 20, 2013||FPAY||Fee payment|
Year of fee payment: 8